Introduction

Recent advancements in designer DNA editors have demonstrated significant therapeutic potential for treating genetic disorders1,2,3,4,5. However, for the successful clinical translation of these tools in cell and gene therapies, a comprehensive understanding of their cellular activities is essential to enhance editing efficiency and ensure safety. Consequently, numerous techniques have emerged to detect or anticipate genotoxic events, reflecting an increasing demand for thorough characterisation of these genome-editing tools. Despite this progress, these detection methods present several limitations, including high costs, time-consuming protocols, the necessity for extensive bioinformatics expertise, and the risk of selective and sometimes biased evaluations of aberrations. Furthermore, the temporal dynamics of DNA editing and the subsequent repair processes remain poorly understood, leading to potential misinterpretations and observational biases.

Double-strand breaks (DSBs) are one of the most critical DNA lesions induced during genome editing and are typically resolved through non-homologous end-joining (NHEJ), alternative end-joining (alt-EJ), or homology-directed repair (HDR) pathways6. Programmable endonucleases, such as the clustered regularly interspaced short palindromic repeats-associated protein 9 (CRISPR-Cas9), leverage these repair mechanisms to achieve gene knockouts or, when combined with donor templates, facilitate the insertion of customised sequences into specific loci7,8,9. Given the potential for long-term therapeutic effects, targeted genome editing offers promising applications for a broad spectrum of genetic diseases10. However, unintended repair outcomes, such as indels, large deletions, and other substantial chromosomal aberrations, often hinder the desired editing effects11,12,13,14 and pose significant genotoxic risks that could compromise both the safety and efficacy of these therapies15.

To address these challenges, targeted integration enhancers (TIEs) have been developed by repurposing NHEJ and microhomology-mediated end-joining (MMEJ) inhibitors, such as AZD7648 and ART558, respectively, to promote template-mediated homology directed repair (HDR)16,[17](https://www.nature.com/articles/s41467-025-65182-4#ref-CR17 “Selvaraj, S. et al. High-efficiency transgene integration by homology-directed repair in human primary cells using DNA-PKcs inhibition. Nat. Biotechnol. 1–14 https://doi.org/10.1038/s41587-023-01888-4

(2023).“). While TIEs represent a promising strategy to minimise undesirable on-target DSB repair products, there is still a considerable lack of information regarding their impact on both on- and off-target aberrations.

Although many techniques have been developed to detect and quantify specific designer nuclease-induced aberrations, they fail to provide an absolute assessment of the frequency at which all undesired aberrations occur at on- and off- target sites. To quantify editing efficiency, most conventional methods rely on target site PCR amplification, such as sequencing based methods (e.g., ICE18, CRISPRESSO19), enzymatic (e.g. T7E120), or others (e.g. IDAA21; dPCR22, rhAmpSeq23,24). While these strategies are effective for detecting small indel populations, they do not identify large deletions, DSBs, or other complex aberrations due to the inability of these sequences to be PCR amplified. Long-read sequencing offers greater detection capabilities for large deletions, but comes with higher costs, demands for specialised protocols and bioinformatic knowledge, and is still biased by kilobase spanning deletions25,26,[27](https://www.nature.com/articles/s41467-025-65182-4#ref-CR27 “Hwang, G.-H. et al. Large DNA deletions occur during DNA repair at 20-fold lower frequency for base editors and prime editors than for Cas9 nucleases. Nat. Biomed. Eng. 1–14 https://doi.org/10.1038/s41551-024-01277-5

(2024).“). These challenges are similar across various high-throughput, PCR-based techniques used to detect translocations (e.g., CAST-seq28, HTGTS29, UDITAS30) or genome-wide off-target effects (e.g., GUIDE31, CHANGE32, DIGENOME33, BLISS34, qDSB35, DISCOVER-seq36). Alternatives that provide broader genomic characterisation, such as whole-genome sequencing, G-banding, FISH karyotyping, or comparative genomic hybridisation, sacrifice detection sensitivity for a wider overview37. Recent advances in imaging have led to the development of optical genome mapping38, which bridges the resolution gap between karyotyping and NGS techniques on the kilobase scale (Supplementary Fig. 1a). By comparing the outcomes from various techniques, a clearer understanding of the mutational landscape can be achieved, although integrating these results remains challenging. Consequentially, the rate of endonuclease cleavage and subsequent DSB repair remain poorly characterised across cell types and gene targets due to the lack of a highly sensitive and quantitative assessment strategy28,31,32,39,40. Quantifying DSB induction over time has previously relied on partial or indirect measurements, such as repair product modelling from amplicon sequencing41,42,43 or immunoprecipitation44. Furthermore, existing strategies have likely underestimated the frequency of precise DNA repair, and the time required for a cell to resolve a DSB. A thorough understanding of DNA cleavage and repair kinetics is essential for generating the safest and most effective gene editing platform for clinical purposes.

To measure fundamental editing processes, we develop a comprehensive assembly of multiplexed digital PCR (dPCR) assays termed CLEAR-time dPCR (Cleavage andLesionE**valuation via AbsoluteReal-Time dPCR), a method previously referred to as MEGA dPCR (Multipurpose Editing and Genotoxicity Assessment), which reveals the prevalence of scarless DNA repair after DSBs, which subsequently leads to recurrent cleavage events by nucleases via dual normalisation after dPCR analysis.

CLEAR-time dPCR addresses critical gaps in the available genetic engineering analysis toolkit by providing a rapid, accessible, and specific overview of genome integrity post-gene editing that can be applied effortlessly to clinically relevant samples.

Results

CLEAR-time dPCR reliably quantifies nuclease-induced aberrations

By taking advantage of the ability of dPCR to quantify the absolute number of copies and linkage of DNA molecules45,46, we developed a modular ensemble of multiplexed dPCR assays to quantify the integrity status of the DNA and its repair outcomes following delivery of CRISPR-Cas ribonucleoprotein (RNP) reagents to human primary CD34+ HSPCs (Fig. 1a).

Fig. 1: CLEAR-time digital PCR reliably quantifies nuclease-induced aberrations in HSPCs.

a Diagram illustrating the induced aberrations by designer nucleases and the CLEAR-time dPCR assay strategies. b Absolute copy number and linkage normalisation workflow. Data are shown as mean ± s.d.c Genome copy frequency summarisation of on-target aberrations 3 h, 3- and 14-days post-Cas9 editing with and without AAV transduction targeting the WAS locus. Data are shown as mean ± s.d. d Single cleavage restriction digestion of AAV donor template VCN and % integrated donor template measured in AAV-transduced cells 3- and 14-days post-editing. Data are shown as mean ± s.d.e End trimming was measured as absolute loss of 5’- and 3’- sequences flanking the Cas9 cleavage site 3 h, 3- and 14-days post-Cas9 editing in RNP-only edited cells. Data are shown as mean ± s.d.f Aneuploidy measured as the absolute change of p or q arm copy numbers. Data are shown as mean ± s.d. g Validation of indel frequency by comparing the relative indel frequency calculated by dPCR, T7EI assay, ICE and NGS measured in WAS edited HSPCs 3-days post-editing. Abs. and Rel. refer to absolute and relative indels (i.e., normalised or not normalised to a reference), respectively. Data for T7EI, ICE, and NGS represent n = 1, dPCR data shown as mean ± s.d. of n = 3 technical replicates. h Validation of donor-template integration by comparing digital PCR to flow cytometry of AAV-transduced cells at 3- and 14-days post-editing. Flow cytometry data represent n = 1, dPCR data shown as mean ± s.d. of n = 3 technical replicates. i Qualitative validation of large deletion and other aberrations of RNP only edited cells 3-days post-editing using CAST-seq. On top, WAS gene schematic, exons in bold. Light blue indicates aberrations on negative-strand, light red indicates aberrations on positive-strand (n = 2 technical replicates). j NGS targeted sequencing spanning ~2500 bp of the cleavage site targeting WAS (white arrowhead) indicating small and large deletions ( > 250 bp). X-axis indicates nucleotide position; Y-axis indicates number of mapped reads. Scale bar indicates 250 bp. All data represents n = 3 technical replicates unless stated otherwise. b, c, e One-way ANOVA with Sidak’s multiple comparison test. d Two-way ANOVA with Tukey multiple comparisons test. f Two-way ANOVA with Sidak’s multiple comparison test. n.s.= no statistical significance, ****p < 0.0001. Source data are provided as a Source Data file.

Wildtype, indels, and total-non-indel aberrations (Edge assay)

To quantify the fraction of loci harbouring wildtype sequences, indels, and total non-indel aberrations, an “Edge” assay comprised of a single pair of primers placed on either side of the RNP target site is used. This assay is complemented with a “cleavage” fluorescent-labelled oligo, i.e., a FAM probe placed directly over the prospective cleavage site, and a “distal” fluorescent-labelled oligo, i.e., a HEX probe placed ~25 bp up- or down-stream of the FAM probe (Fig. 1a; i). We observed a clear FAM/HEX double-positive population of droplets in mock electroporated cells, where no on-target cleavage (and therefore no mutations) was expected. In RNP-electroporated cells, the FAM cleavage probe does not efficiently bind to mutated amplicon sequences, resulting in an attenuated or total loss of FAM signal and, consequently, a drop in the number of amplified copies, thereby indicating indels. Non-indel aberrations such as large deletions, translocations, and unresolved DSBs at the on-target cleavage site completely disrupt amplification, resulting in a total loss of both the FAM and HEX copies relative to the reference assay (Fig. 1a; i, Fig. 1b; Supplementary Fig. 1b; i). The sum of Wildtype, indel, and other aberration populations encapsulates the entire edited cell population.

Double-strand breaks, large deletions, and other structural mutations (Flanking and linkage assay)

The “Flanking” assay quantifies double-strand breaks, end processing, large deletions and other aberrations. It involves two amplicons flanking the cleavage site (5’ and 3’), each with a probe nested within the primer pairs (Fig. 1a; ii). The linkage between these probed sequences is measured by the presence of double-positive signals within the same PCR droplet, normalised against the frequency expected by chance, as previously described46. In unedited cells, the Flanking assay predominantly yielded double-positive droplets, indicating minimal random DNA fragmentation from the genomic DNA isolation. Conversely, an increase in single-fluorescent positive droplets was observed in samples from RNP-electroporated cells, reflecting a decrease in linked sequences due to DSBs. Additionally, when the end processing involves both DNA strands on one or both sides of the cleavage, it prevents primer/probe binding, resulting in a loss of copies (Fig. 1b, Supplementary Fig. 1b; ii). It is important to note that the classification of large deletions is determined by the distance between the primer and probe relative to the cleavage site. Therefore, any DNA end processing greater than 20-30 bp from the cleavage site is classified as a large deletion, depending on the specific design of the assay.

Aneuploidy assay

Aneuploidy, i.e., full, or partial loss/gain of chromosomes, was quantified in this assay. Numerical variation of the p arm, q arm, or the whole chromosome results in the alteration of FAM, HEX, or both signals, whereas balanced segregated translocations could be observed as a gain of these signals. To quantify this, primers and probes are placed in the sub-telomeric regions of the p and q arms of the edited chromosome (Fig. 1a; iii, Fig.1b; Supplementary Fig. 1b; iii).

Target-integrated and off target-integrated/episomal donor assessment

The frequency of integrated and non-integrated templates when using a donor sequence for DSB correction via HDR was calculated using the “Targeted integration and episomal” assay. The assay consists of a primer bound to genomic sequence outside of the regions complementary to the donor homology arms, and a donor-specific primer, and a probe between them. A second set of primers is placed within the donor-specific sequence to detect all donor template copies, i.e., on- or off-target integrated and non-integrated episomes (Fig. 1a; iv, Fig. 1b; Supplementary Fig. 1b; iv, v). To evaluate donor DNA concatemers, an enzymatic digestion was performed within the donor sequence but outside the region covered by the second set of primers.

Reference assays

A pair of primers and probes on non-targeted chromosomes placed equidistant from those used in the flanking assays acts as a loading control and allows for copy and linkage normalisation to obtain an unbiased quantification of mutations identified by the mutation detecting assays (Fig. 1a; v, Fig. 1b; Supplementary Fig. 1b; vi).

Additional assays

CLEAR-time dPCR can integrate various dPCR assays to enhance the detection of specific mutations around the target site. By designing custom sets of primers and probes based on knowledge gathered from sequencing-based detection techniques, it is possible to improve discrimination of particular mutations, such as microhomology-mediated end-joining (MMEJ) repair, translocations, or inversions. This not only enriches CLEAR-time dPCR’s mutation discrimination capacity but also maintains the absolute character of the assessment, seamlessly integrating with the detection of other DSB repair products and reducing the quantity of mutations grouped as ‘other aberrations’.

Double normalisation

A distinguishing feature of this methodology is the double normalisation of copy number and linkage, eliminating the biases derived from a relative or partial quantification (Supplementary Information 1). Briefly, each assay copy number is first normalised against averaged reference assays values. Subsequently, this initial ratio is normalised again, using values from the control unedited sample. This second normalisation accounts for inter-assay variabilities and accurately measures the effective variation, allowing quantification of absolute mutation population frequency (Supplementary Fig. 1c).

Dynamic and unbiased chromosomal aberration analysis and validation

To demonstrate the capabilities of CLEAR-time dPCR, primary human HSPCs were electroporated with a SpCas9/sgRNA RNP targeting the therapeutically relevant X-linked WAS gene, encoding the Wiskott-Aldrich syndrome protein. RNP electroporated HSPCs were split and either left as an RNP-only treatment or, by AAV transduction, delivered a donor-template encoding GFP flanked by 700 bp arms homologous to the RNP cleavage site7. Mock-electroporated cells were used as a control. Viable cells were counted (Supplementary Fig. 1d), and genomic DNA was then extracted 3 h, 3 days, and 14 days post-editing and analysed using CLEAR-time dPCR.

By applying the normalisations and summing the absolute locus frequencies, we created a unified plot for each condition. It is important to note that the retrieved data may not contain 100% of copies due to an accumulation of errors that exceed the logical limit. In this scenario, a careful review of setting dPCR droplet thresholds, an increased number of reference assays, or the exclusion of clear copy number outlier replicates may help reduce the error. Alternatively, an unexpected donor recombination and duplication or unusual structural rearrangements may be at the base of a biological reason.

Three h post-editing, the majority of other aberrations (68.4 ± 0.4%) were derived from DSBs with early evidence of indels and large deletions (Fig. 1c). After 3 days, DSBs mostly resolved as indels, though a small percentage of other aberrations persisted, likely representing a combination of unresolved DSBs and other gross chromosomal aberrations such as translocations or inversions. In AAV-transduced cells, donor-template integration is preferred mostly at the expense of indels, as observed by a significant decrease in the absolute percentage of indels between the RNP and RNP + AAV treatment groups (67.7 ± 1.9% and 38.4 ± 0.4%, respectively) (Fig. 1c). AAV-transduced cells also showed fewer large deletions and other aberrations compared to the RNP-only treatment group, albeit to a lesser extent than indels in absolute terms (Fig. 1c). Although large deletions remained stable at all time points post editing, the frequency of DSBs continued to drop to 7.0 ± 0.2%, indicating a continuous dynamic between DNA repair, loss of detrimental aberrations, and nuclease cleavage. The constant amount of wildtype loci would also reflect an ongoing nuclease activity 3 days post-editing and mainly later resolved as indels (Fig. 1c).

As the homology arms of the AAV donor-template are synonymous with the on-target WAS sequence, the flanking assay is unable to differentiate between large deletions and other aberrations when episomal AAV is present. However, with the TI assay we could instead track GFP copies in AAV-transduced cells 3 to 14 days post-editing (Fig. 1d). Although there was a 2-fold reduction of GFP copies between the two timepoints, the majority of the donor template molecules were still detectable ( > 99.4%), likely in the form of non-integrated episomes, and an even smaller amount likely integrated at off-target loci, as observed previously47,48. Additionally, digestion with a restriction enzyme with activity at a single cleavage site within potential concatemerized donor sequences did not result in a significant increase in VCN, suggesting these structures had not efficiently formed in the edited cells (Fig. 1d). DNA trimming of the 5’ and 3’ sequences flanking the cleavage site was dynamic, peaking at 3 days post editing at 14.6 ± 2.6% and 12.8 ± 2.6% for the 5’ and 3’ flanking sequences, respectively; however, there was no trimming bias in a particular direction (Fig. 1e). Conversely, there were no significant copy number deviations in chromosome sub-telomeric regions (Fig. 1f), indicating minimal aneuploidy.

To validate CLEAR-time dPCR, we aimed to benchmark and quantify double-strand breaks, indels, targeted integration, and chromosomal aberrations using orthogonal techniques. Digital PCR has been used to quantify double-strand breaks previously43, however, this particular strategy results in the over estimation of DSBs due to the miscategorising of other aberrations, such as large deletions (Supplementary Fig. 1e).

The detection of indels demonstrated concordance across multiple techniques, including non-reference normalised dPCR, sanger and next-generation sequencing (NGS), and the T7EI assay (Fig. 1g; Supplementary Fig. 1f). However, since these methods are inherently biased, amplifying primarily intact sequences with few or no base insertions or deletions, CLEAR-time dPCR suggests that they overestimate indel frequency by approximately 17% in these samples. To verify donor template integration, we performed flow cytometry on AAV-transduced cells and found that 45.3% of cells were GFP positive, which was comparable to the 53.4 ± 1.01% calculated by the targeted integration dPCR assay (Fig. 1h).

Lastly, to confirm the presence of large deletions and gross chromosomal aberrations, we employed CAST-Seq and 5 kb amplicon sequencing. Around the cleavage site, we identified inverted sequences and kilobase-spanning large deletions (Fig. 1i, j), providing qualitative diagnostic evidence for such events. The quantification of these events, as performed in previous publications[49](https://www.nature.com/articles/s41467-025-65182-4#ref-CR49 “Cullot, G. et al. Genome editing with the HDR-enhancing DNA-PKcs inhibitor AZD7648 causes large-scale genomic alterations. Nat. Biotechnol. 1–5 https://doi.org/10.1038/s41587-024-02488-6

(2024).“), can suffer from significant biases due to larger deletions exceeding the distance of placed primers and gDNA extraction quality, thus skewing the interpretation. However, the qualitative approach used here complements the robust quantification achieved with CLEAR-time dPCR.

Taken together, these data show that the combined dPCR assays and appropriate normalisations provide a comprehensive, robust, and validated overview of the heterogeneous DSB repair products within the total post-editing cell population.

CLEAR-time dPCR quantifies chromosomal aberrations and cell clonal dynamics in clinically relevant targets

We next sought to quantify the sensitivity of CLEAR-time dPCR and to demonstrate its broad-ranging applications. Clonal proliferation of cancerous cells from genotoxic by-products of gene therapy is a legitimate safety concern50. CLEAR-time dPCR can diagnose variations in the frequency of specific mutations within the edited loci at early stages, allowing for the prediction and tracking of cell clonal expansions. To determine the limit of detection (LoD) of a mutation within a population of wildtype sequences, we artificially recreated a large deletion event at the CCR5 target sequence, which occurs in HSPCs after Cas9 editing (Fig. 2a). We synthesised two reference linked cassette oligonucleotides: a “wildtype” and a 5’ end truncated “large deletion” sequence mimicking that observed in the CCR5 gene (Fig. 2a; Supplementary Fig. 2a). By mixing the cassettes in decreasing “large deletion” to “wildtype” ratios and performing the dPCR flanking assay, we determined the LoD with this assay to be 2.33% (Fig. 2b).

Fig. 2: CLEAR-time dPCR detects chromosomal aberrations and cell clonal dynamics in clinically relevant targets in vitro and in vivo.

a Schematic of wildtype and large deletion cassettes used to establish the LoD of large deletions, and representation of clonal expansion of cells harbouring a large deletion (orange) mediated genotoxic aberration amongst wildtype cells (blue). Dark and light grey bar indicates CCR5 and reference sequences, respectively. Red bar indicates flag sequence used to fuse assay and reference sequences. b Correlation of observed against the expected percentage of large deletions. Vertical black dotted line represents the limit of detection. Solid and dotted red line represents line of regression with 95% confidence interval, respectively. R2 = 0.997. Calculated with linear regression analysis in GraphPad. Data shown as mean ± s.d. of n = 4 technical replicates. c CLEAR-time dPCR summaries of Cas9-edited HSPCs targeting various genes at different timepoints. The BTK edited HSPCs were also transduced with AAV6 encoding GFP. All editing was normalised against unedited mock electroporated HSPCs. Data are shown as mean ± s.d. d CLEAR-time dPCR summary of SH2D1A edited T cells with Cas9, Cas12, and TALENs at 3 days post-editing. Data are shown as mean ± s.d. e CLEAR-time dPCR summary of on-target CCR5 edited with decreasing concentrations of Cas9 at 3 h and 3 days post-editing. Data are shown as mean ± s.d. f DSB and indels quantification at three known off-targets targeting CCR5 with decreasing concentrations of Cas9 at 3 h and 3 days post-editing. Data are shown as mean ± s.d. g CLEAR-time dPCR on XIAP edited HSPCs pre-transplant and 16 weeks post-transplant. Data are shown as mean ± s.d. n = 4 mice for each treatment group. h Integration frequency in 16-week post-transplant XIAP edited and AAV-transduced hCD45 cells by dPCR and flow cytometry. Flow cytometry data represent n = 1 per mouse, dPCR data shown as mean ± s.d. of n = 3 technical replicates. i CLEAR-time dPCR normalised ICE analysis of pre-transplant XIAP edited HSPCs and post-transplant XIAP edited hCD45 cells. All data represents n = 3 technical replicates unless stated otherwise. d–f Two-way ANOVA with Tukey post-hoc test. n.s.= no statistical significance, ****p < 0.0001. Source data are provided as a Source Data file.

To demonstrate the flexibility of CLEAR-time dPCR in evaluating the loci status of different gene targets, we electroporated HSPCs with RNP complexes targeting six therapeutically relevant genes with previously published gRNAs and tracked the consequences of DSB repair at these loci over time. To assess the contribution of HDR to loci repair, we also delivered a donor-template by AAV transduction following RNP electroporation for the BTK gene target (Fig. 2c; Supplementary Fig. 2b). Unlike the stable integration of the donor cassette via HDR in the WAS locus observed in AAV-transduced cells across timepoints, site-specific integration events at the BTK locus dropped by almost 2-fold from early to late timepoints and were replaced by indels (Fig. 2c; Supplementary Fig. 2c). The frequency of DSBs observed at 3 h post editing across the targets varied greatly from 18.5 ± 0.5% to 55.0 ± 0.4% (Fig. 2c). Indels marginally increased as other aberrations decreased between the 3-day and later timepoints (Supplementary Fig. 2d), whereas DNA end-trimming typically peaked at day 3 and different targets showed significant aneuploidy (Supplementary Fig. 2e).

It is noteworthy that across the various sites examined in this study, CCR5 (also considered a “safe harbour locus“51), EMX1, FANCF, RAG1, VEGFA, IL7R, CD34, and WAS (with the exception of BTK edited cells without a donor), we observe a stable or decreasing number of WT sequences while indels continue to increase beyond 72 h. Typically, DSBs delay cell proliferation through p53/p21 signalling52. Thus, cells exhibiting absent or reduced cleavage efficiency should theoretically possess a proliferative advantage, resulting in an increase in WT alleles over time, contrary to our observations.

These data strongly support the hypothesis of protracted nuclease activity, albeit significantly diminished, beyond 72 h post-electroporation.

An alternative hypothesis might consider the creation of mutations that confer a proliferative advantage. However, certain genes edited in this work, such as WAS, BTK, and FANCF, would rather hinder growth when mutated53,54,55. Furthermore, other sites examined, like VEGFA (which enhances vascularisation in tumours) and RAG1 and FANCF (which contribute to genomic instability), have not been previously associated with short-term proliferative advantages in vitro, unlike p53 or p21 knockouts.

Consequently, the most plausible explanation for these observations remains the persistent activity of the nuclease.

Notably, end-trimmed events, which consistently occur alongside indels, tended to disappear from the overall population over time, indicating that those mutations are associated with non-replication permissive aberrations or that the sequence is restored via homologous recombination with the sister chromatid as template.

CLEAR-time dPCR detects aberrations across various designer nucleases and fine-tunes editor activity

To explore whether CLEAR-time dPCR is applicable across different cell types and designer nucleases, we electroporated PBMC-derived T cells with Cas9, Cas12a, or TALEN, each targeting the same position in the SH2D1A gene that encodes Slam-associated protein (SAP)56. Due to subtle variations in the precise cleavage site across the tested nucleases, we ensured that the 5’ end of the FAM cleavage probe was located over the nucleotides most likely to form indels, as probe placement is integral for discriminating small indels (i.e., +1 insertions) from wildtype alleles (Supplementary Fig. 2f). We found a similar frequency of all types of aberrations induced by the different nucleases, again with no significant evidence of aneuploidy; however, cells edited with TALEN did show a 3’ directional processing bias (Fig. 2d; Supplementary Fig. 2g). Amplicon sequencing revealed that only Cas9 produced +1 indels, which were clearly visible as a distinct droplet population in the dPCR dot-plot (Supplementary Fig. 2h).

Mitigation of the off-target cleavage activity of an RNP targeting the CCR5 gene when using a high-fidelity Cas9 has previously been described28.To study the effect of Cas9 concentration on on/off-target cleavage activity, we edited cells at the CCR5 locus with decreasing concentrations of RNP and performed CLEAR-time dPCR at the cut site on chromosome 3 as well as at off-target sites on chromosomes 1, 13 and 19. At 3 h, we observed a clear pattern of decreasing DSB generation for both the on and off-target locus correlating with decreasing amounts of RNP, though notably we observe a significant decrease in off-target activity when reducing the RNP concentration to 1/8th of the original amount (Fig. 2e, f, Supplementary Fig. 2i). This pattern was observed again regarding indels three days post-editing; however, non-indel aberrations at the on-target site were present until the lowest RNP concentration (Fig. 2e, f). These results demonstrated the effective utilisation of CLEAR-time dPCR to determine the optimal nuclease concentration for targeted cells, thereby mitigating unwanted effects and maintaining high on-target efficiency.

HSPC aberration landscape follow-up in longitudinal murine studies

Recent pre-clinical studies have consistently observed a significant decrease in genome-edited HSPCs harbouring the intended therapeutic sequence post-transplantation57. We corroborated this phenomenon by screening gene edited transplanted cells with CLEAR-time dPCR and further characterised the mutation dynamics occurring after xenotransplantation. We obtained human HSPCs in which the gene encoding X-linked inhibitor of apoptosis protein (XIAP) had been targeted with SpCas9, transduced with an AAV encoding either a GFP- or codon-optimised XIAP sequence or GFP with homology arms synonymous with the target site, and transplanted into immunodeficient NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ NSG mice (Fig. 2g).

CLEAR-time dPCR revealed that prior to transplantation in AAV-transduced cells, indels were again found to be reduced at the expense of integration, with an integration frequency in AAV-transduced cells found to be ~25% (Fig. 2g; Pre-transplant). Sixteen weeks post-transplantation, the aberration frequencies showed a large degree of variability, with evidence of skewed dominance of mutant sequences (Fig. 2g; Post-transplant). As expected, alleles carrying the template sequence were significantly reduced post-transplantation. This was confirmed by flow cytometry (Fig. 2h; Supplementary Fig. 2j). To resolve whether there was clonal proliferation of a particular indel repair product, we sequenced the post-transplant samples and normalised the absolute population frequencies using CLEAR-time dPCR. We found an overrepresentation of +1 and −2 indels compared to the cells initially transplanted into the mice (Fig. 2i). Additionally, cells derived from two mice that received HSPCs edited with the GFP control vector showed a large increase in non-indel aberrations as opposed to indels (Fig. 2i). Together, these data illustrate that CLEAR-time dPCR has valuable roles in all stages of gene therapy development, from early therapy development to pre-clinical models and post-clinical follow-up once administered to patients and combined with orthogonal methods will further the characterisation in the repair outcomes.

Targeted integration enhancers decrease genome stability in HSPCs, iPSCs, and T cells

Integrating a template sequence at a specific location using HDR for gene therapy has significant advantages, but its effectiveness is hindered by indel formation through the NHEJ or Alt-EJ repair pathways, which are semi-dependent on the cell cycle58,59,[60](https://www.nature.com/articles/s41467-025-65182-4#ref-CR60 “Brambati, A. et al. RHINO directs MMEJ to repair DNA breaks in mitosis. Science 381, 653–660 (202